The AWS Big Red Button Is Finally Here
Before transitioning to cloud engineering, I worked as an electrical engineer. In any industrial setting, one of the first things you learn about is the “big red button”: the Emergency Power Off (EPO). If things go sideways, you press it, and everything shuts down instantly.
When I started tinkering with AWS while studying for my Solutions Architect certification, I brought that mindset to the cloud. I thought I had cleaned up everything after a lab session, but a few hours later, I received a budget alert showing I had crossed my $0 spend threshold.
That’s when the panic set in. I'd left a NAT Gateway running, and as the hours ticked toward 21, the bill kept climbing. In that exact moment, I desperately wished for that “big red button” to freeze everything. With no credits left to cover the cost, I was stuck. I ended up posting my horror story on Reddit, and luckily, the AWS Support team spotted it and saved the day.
The great news is that the big red button is here and at no additional cost!!!
This feature is currently available to a limited number of customers.
The current prerequisites are:
(1) It is currently limited to five services: EC2, RDS, Lambda, Bedrock, and SageMaker.
(2) You must be on a paid plan to create a spend limit.
(3) Spend Limit works on a Project Basis
(4) $20 is the baseline hardcoded into the system as the absolute global minimum (Even if you have a brand-new, empty AWS account with $0 in spending and zero active resources running, AWS will still not let you set a spend limit lower than $20) [1].
AWS recommends that the Spend limit be used on experimentation, learning, and Sandbox workloads. It can be used for Production environments but requires extra caution because you don’t want to interfere with workloads that serve customers, leading to downtime and loss of revenue. Moreover, it is useful for those workloads that can tolerate a brief pause when unexpected costs are incurred.
Spend Limits work per project in helping teams operate within a predictable and controlled budget per month. Thus, if your designated project reaches its limit, AWS automatically pauses the project, meaning it stops its resources so that the costs remain within the stipulated budget [2].
The Rule of Two: How AWS Finds the Floor
The spend limit is the ceiling on the project’s pre-tax costs. When you set a spend limit, AWS automatically calculates the absolute lowest number you are allowed to enter, and it chooses whichever is greater between:
1. A flat $20 baseline.
2. A conservative estimate of what you are actually on track to spend.
The conservative estimate isn’t arrived at by using a simple math projection (like multiplying your daily spend by 30). Rather, it looks at the active resources (i.e., what is running now, your month-to-date spending, and last month’s total bill, and if last month was high, then that becomes the baseline.
The Formula in Action: A Real-World Example
Let’s see how this works in practice. Imagine it’s October 15th, and you are reactivating a project that you had paused for a few weeks. You want to set a strict spend limit. Here is what your account looks like:
- Spend so far this month: $15 (from the days before you paused it)
- Uncharged usage:
$5 (pending costs not yet on the invoice)
- Last month’s bill (September): $80
- Current active resources: You just spun up a few servers that cost $4 per day to run. With 16 days left in October, you are on track to spend another $64 ($4 × 16 days) if you leave them on.
Step 1: AWS Calculates Your “Conservative Estimate”
Get Moses Kuria’s stories in your inbox
Join Medium for free to get updates from this writer.
AWS tallies your real-time footprint for the current month:
$15 (spent) + $5 (uncharged) + $64 (projected for active resources) = $84
AWS then compares this $84 projection against last month’s bill ($80). Because $84 is higher, $84 becomes your conservative estimate.
Step 2: AWS Determines Your Minimum Allowed Limit
Finally, AWS compares its estimate against the flat $20 baseline:
- Option A: $20
- Option B: $84
AWS takes the larger number. Therefore, your minimum allowed spend limit is $84.
The Takeaway: How to Lower Your Limit
If you look at that $84 floor and think, “No way, my budget is only $40,” you can’t just type in a lower number. AWS won’t let you.
Notifications are sent out when the actual costs reach 50%, 75%, and 90% of your limit or when on track to reach or exceed your spend limit in the next 10 days.
Customizing AWS Spend Limits & Early Cost Controls
There are several ways to keep the bill from getting past the amount set.
1. Stop resource creation — stops the creation of resources 7 days before reaching the limit, where new resources are stopped from launching. AWS managed SCPs are implemented to stop the services from launching and restrict them to only that project and not services outside the project. The resources running stay running, but in instances where autoscaling is used, they might be disrupted.
2. Pause idle resources — 5 days before reaching the limit, all idle resources are paused, with the active ones continuing to run. Pausing is determined by AWS Compute Optimizer. One can view the idle resource recommendations by logging into the project and looking at the saving oppornunities.
3. Why we pause an EC2 instance — EC2 instances with less than 5% CPU utilization and network I/O less than 5MB per day for the past 14 days are paused. The G and P instance types have a different idle criterion [3].
4. SageMaker endpoints are paused if they have not received any invocations in the past 14 days.
5. MySQL and PostgreSQL RDS are paused if they are not a read replica and have not received any DB connections, have had low CPU usage, and have had low read and write activity in the past 14 days.
